In order to precisely predict the hazard degree of goaf (HDG), the RS-TOPSIS model was built based on the results of expert investigation. To evaluate the HDG in the underground mine, five structure size factors, i.e. goaf span, exposed area, goaf height, goaf depth, and pillar width, were selected as the evaluation indexes. And based on rough dependability in rough set (RS) theory, the weights of evaluation indexes were identified by calculating rough dependability between evaluation indexes and evaluation results. Fourty goafs in some mines of west China, whose indexes parameters were measured by cavity monitoring system (CMS), were taken as evaluation objects. In addition, the characteristic parameters of five grades’ typical goafs were built according to the interval limits value of single index evaluation. Then, using the technique for order preference by similarity to ideal solution (TOPSIS), five-category classification of HDG was realized based on closeness degree, and the HDG was also identified. Results show that the five-category identification of mine goafs could be realized by RS-TOPSIS method, based on the structure-scale-effect. The classification results are consistent with those of numerical simulation based on stress and displacement, while the coincidence rate is up to 92.5%. Furthermore, the results are more conservative to safety evaluation than numerical simulation, thus demonstrating that the proposed method is more easier, reasonable and more definite for HDG identification.